Browsing by Person "Nagel-Held, Johannes Henrich"

Now showing 1 - 1 of 1

From spectra to traits: advancements in predicting wheat characteristics
(2025) Nagel-Held, Johannes Henrich; Hitzmann, Bernd
This thesis is driven by two primary objectives: To evaluate the potential for improving quality assessment throughout the wheat value chain and the potential for replacing protein content as the primary determinant of price and quality with spectroscopic methods. An early, fast, and well-founded decision on wheat quality leads to improved processing properties and product quality of wheat products. Given that these attributes are often more closely tied to protein quality or other components than protein quantity, a direct and precise quantification is highly desirable. The noteworthy advantage of protein content lies in its prediction through near-infrared spectroscopy (NIRS). To augment the information available from near-infrared spectra, which are inherently limited in their insights into molecular compounds within samples, NIRS was supplemented with Raman and fluorescence spectroscopy. The thesis`s foundation comprises four distinct sample sets, encompassing both common wheat and spelt. In total, a staggering 4,237 samples of whole grain, whole grain flour, and extracted flour were analyzed to predict a wide array of 100 diverse quality parameters, spanning from measures of protein quality to dough rheological properties, baking behavior, physical and chemical characteristics and agronomic traits. To enable accurate predictions several spectra pre-processing and regression techniques were applied. Among the spectrometers and algorithms tested, no clear recommendation can be given. In this work, NIRS was found to perform well over different levels of sample preparation, while Raman and fluorescence spectroscopy performed better on flour. A glimpse into the findings is best exemplified by considering loaf volume, a measure of end-product quality. Prediction errors (RMSECV) ranged between 29 and 43 mL/100 g of flour, accompanied by R² values spanning 0.66 to 0.78. One major factor influencing the improvable prediction accuracy is the inherent measurement error associated with the baking trial. However, when this measurement error is known, a straightforward solution emerges: By measuring a few samples repeatedly, the prediction error can be corrected from 43 mL/100 g of flour to 28 mL/100 g, aligning with the magnitude of the measurement error itself. Simultaneously, the corrected R² value improves from 0.66 to 0.86. It has been shown that measurement error correction works for other quality characteristics, such as water absorption or baking loss, and may be applied to other parameters. The robustness of the models was evaluated by successfully predicting complex traits of unknown cultivars. However, the models proved less effective in predicting across separate locations and years. Similar robustness results were found for protein content, water absorption and grain yield. Since protein content is a well-established and predictable parameter, robust models for complex traits are also feasible. Other well predictable parameters are protein content (R² = 0.97-0.98, RMSECV = 0.16-0.20 %), wet gluten content (R² = 0.89-0.98, RMSECV = 0.80-0.84 %), water absorption (R² = 0.68-0.83, RMSECV = 0.9-1.2 mL/100 g) and grain yield (R² = 0.73-0.85, RMSECV = 5.9-6.8 dt/ha). Examples for non-predictable parameters are starch properties measured with a rapid visco analyzer or enzyme activity. And for some parameters no clear conclusion could be drawn, such as SDS sedimentation volume, plant height, thousand kernel weight, extensograph or farinograph parameters. The successful prediction of intricate traits can be further refined by reducing measurement errors of the reference values and using better suited spectrometers. The robustness can be increased by expanding the sample set by including data from a broader spectrum of locations and years.Implementing this proposed methodology holds the potential to instigate positive changes within the wheat supply chain, with implications for product quality, environment, and economic aspects.